Introduction
Natural Language Processing of Amazon Reviews
Background
This storyboard aims to display the results of an analysis using natural language processing to analyze Amazon electrical product reviews. We’ll be comparing lexical diversity and lexical density of reviews with 1-5 star ratings, investigating the rough importance of words used in reviews of different ratings using TF-IDF analysis, and finish with a rough sentiment analysis using both the AFINN and BING lexicons.
Much of this analysis is shamefully duplicated from this fantasic article I came accross a few years ago by Debbie Liske which aimed to use natural language processing to investigate the lyrics of prince, I’d highly reccommend giving it a read!
Objective
This storyboard aims to use natural language processing to analyze Amazon electrical product reviews.
Data
The dataset used in this analysis can be found here. There are tons of different product categories here, so feel free to try this yourself and perhaps compare the features of different types of reviews. If you use the this data for any purpose please cite them!
Lexical Analysis
Row
Lexical Diversity and Density
Lexical Diversity and Density
Row
Wordcloud for 1 Star Reviews
Important Words
Popular Words
TF-IDF Analysis
TF-IDF Analysis for 1 Star reviews
| overall | word | n | tf | idf | tf_idf |
|---|---|---|---|---|---|
| 1 | product | 7818 | 0.0076728 | 0 | 0 |
| 1 | bought | 7416 | 0.0072782 | 0 | 0 |
| 1 | time | 7196 | 0.0070623 | 0 | 0 |
| 1 | quality | 4686 | 0.0045989 | 0 | 0 |
| 1 | amazon | 4576 | 0.0044910 | 0 | 0 |
| 1 | money | 4443 | 0.0043605 | 0 | 0 |
| 1 | unit | 3817 | 0.0037461 | 0 | 0 |
| 1 | sound | 3508 | 0.0034428 | 0 | 0 |
| 1 | months | 3483 | 0.0034183 | 0 | 0 |
| 1 | return | 3368 | 0.0033054 | 0 | 0 |
| 1 | computer | 3346 | 0.0032838 | 0 | 0 |
| 1 | support | 3265 | 0.0032043 | 0 | 0 |
| 1 | purchased | 3248 | 0.0031877 | 0 | 0 |
| 1 | cable | 3207 | 0.0031474 | 0 | 0 |
| 1 | reviews | 3027 | 0.0029708 | 0 | 0 |
| 1 | fine | 3020 | 0.0029639 | 0 | 0 |
| 1 | price | 2924 | 0.0028697 | 0 | 0 |
| 1 | device | 2905 | 0.0028510 | 0 | 0 |
| 1 | cheap | 2890 | 0.0028363 | 0 | 0 |
| 1 | power | 2776 | 0.0027244 | 0 | 0 |
TF-IDF Analysis for 5 Star reviews
| overall | word | n | tf | idf | tf_idf |
|---|---|---|---|---|---|
| 5 | price | 57166 | 0.0072611 | 0 | 0 |
| 5 | quality | 51079 | 0.0064879 | 0 | 0 |
| 5 | bought | 47779 | 0.0060688 | 0 | 0 |
| 5 | time | 41376 | 0.0052555 | 0 | 0 |
| 5 | easy | 40266 | 0.0051145 | 0 | 0 |
| 5 | product | 39745 | 0.0050483 | 0 | 0 |
| 5 | sound | 35461 | 0.0045042 | 0 | 0 |
| 5 | cable | 34191 | 0.0043429 | 0 | 0 |
| 5 | camera | 34167 | 0.0043398 | 0 | 0 |
| 5 | recommend | 32428 | 0.0041189 | 0 | 0 |
| 5 | nice | 28656 | 0.0036398 | 0 | 0 |
| 5 | computer | 25743 | 0.0032698 | 0 | 0 |
| 5 | love | 24294 | 0.0030858 | 0 | 0 |
| 5 | perfect | 22454 | 0.0028521 | 0 | 0 |
| 5 | purchased | 21782 | 0.0027667 | 0 | 0 |
| 5 | power | 20941 | 0.0026599 | 0 | 0 |
| 5 | amazon | 20837 | 0.0026467 | 0 | 0 |
| 5 | battery | 20382 | 0.0025889 | 0 | 0 |
| 5 | excellent | 20168 | 0.0025617 | 0 | 0 |
| 5 | fine | 20062 | 0.0025482 | 0 | 0 |
Sentiment Anlysis
Row
Chord Plot of Sentiment per Review Score
